software energy efficiency, PMU, DVFS, random-forest, software containers.

The massification of streaming services, online games, network file storage, e-learning, among others applications, imply the demand for high-performance and high-availability computing clouds, which, in turn, require robust, versatile and accessible datacenters. Considering this trend, it is fundamental the development of solutions towards the optimization of the energy consumption in environments with concentrated computational infrastructure, such as datacenters and HPC systems. In this work, a software framework will be presented focused on application performance analysis and energy consumption management. Using orthogonal projection of tensors and regression by Random Forest, an application performance prediction model is developed. Combined with the environment power model, DVFS and software containers, the framework aims at scheduling and real-time adjustment of computational resources and hardware parameters in order to reach the optimal point of energy efficiency of the application, considering QoS constraints.